The last decade has seen a dramatic increase in the development of peer-provided services for adults with serious mental illnesses. The empirical evidence supporting such services lags behind their rapid proliferation, however. Reviews have shown that people with histories of mental illness can provide certain conventional services at least as effectively as non-peers, but have found little evidence for their superiority on a wide range of outcomes. Very few of these studies evaluated peers providing peer support as its own service, though, and even fewer evaluated those aspects of peer support which are thought to be unique to this form of service delivery. The proposed study seeks to be one of the first to use rigorous methods to evaluate the effectiveness peer support as compared to other services offered with the same frequency and intensity by both peers and non-peers. Specifically, the proposed study aims to: 1) evaluate the effectiveness of peer support, as compared to usual care and to an equivalent amount of support offered by peer case managers and non-peer recovery mentors; 2) evaluate the influence of patient diagnosis as a potential moderating variable in determining the effectiveness of peer support; 3) evaluate the cost-effectiveness of peer support; and 4) integrate and test a theoretical model of the active ingredients of peer support. Based on the literature on peer support and the research team's own work in this area, these ingredients are hypothesized to be an instillation of hope through positive self-disclosure; role modeling of self-care and exploring new ways of using experiential knowledge; and a trusting relationship characterized by empathy linked with conditional regard. To achieve these aims, we propose to conduct a mixed-methods, qualitative/quantitative study. The qualitative component will involve interviews with participants both early and late in the study, with the aims being to refine the theoretical model and measures (early) and inform and illustrate the quantitative data analysis (later). For the quantitative component, we propose a randomized, controlled design involving a total of 320 adults with serious mental illnesses and histories of multiple hospitalizations who will be randomly assigned to one of the following conditions: 1) usual care; 2) usual care plus the services of a peer case manager; 3) usual care plus the services of a non-peer recovery mentor; and 4) usual care plus the services of a peer recovery mentor. Fidelity measures will be used to ensure the integrity of the 4 conditions. Data will be collected during the index hospitalization and at 3 and 9 months after discharge. Outcomes will include hospitalization events and days, symptoms, clinical and functional status, community inclusion, substance use, and perceived recovery and quality of life. Changes in demoralization, hope, self-efficacy, self-care, the relationship between participant and case manager or mentor, and social support will also be assessed to test the theoretical model informing the proposed study. Diagnostic data will be collected to determine its influence on effectiveness and cost data will be collected to enable calculations of the cost-effectiveness of peer support.